Professional Documents
Culture Documents
TRITA: IIP-13-01
www.kth.se
Abstract
Manufacturing engineering in general has experienced an increased demand of
process planning in order to optimize processes to reduce costs, environmental
impact and increase time efficiency. Resistance spot welding (RSW) is a common
and large-scale joining method in several manufacturing industries indicating
significant potentials of efficient process planning.
The goal of this thesis is to establish new knowledge for updated and improved
process planning of RSW in industrial applications. The goal is expressed by two
research questions targeting the issues of process variations and potential of
numerical methods of the RSW process. The research questions are expressed in
terms of weld size, which is the main interest in RSW process planning.
The intentional variations of weld sizes due to process parameters, which is the
most important aspect of RSW process planning, have traditionally been analyzed
through physical testing. In the present thesis two numerical methods were
evaluated; regression analysis and FE simulations. For the regression analysis
several models were generated and showed a standard deviation of residuals
between model and physical results of 0.5 mm. For the FE simulations, material
models for the RSW were generated and the simulations showed a standard
deviation compared to physical testing of 0.68 mm. In conclusion, the present
thesis presents results, which help quantify variations in weld sizes and present the
capability of numerical methods of the RSW process.
i
Acknowledgement
This thesis is the result of my research at the department of production
engineering, Royal Institute of Technology, Stockholm within the project SpotLight
between January 2010 and December 2012. The project was funded by VINNOVA
and the project partners; Gestamp Hardtech, Royal Institute of Technology, SAPA,
Scania, SSAB, Swerea IVF, Swerea KIMAB, Volvo Car Group, Volvo Trucks and
University West.
Finally, a special thank you is due to my friends and family, who have supported me
during my research.
Oscar Andersson
ii
Table of contents
1 Introduction.........................................................................................................1
1.1 Background..................................................................................................2
1.2 Outline to the thesis ....................................................................................4
1.3 Research questions......................................................................................4
1.4 Research methodology ................................................................................5
1.5 Introduction to the papers ..........................................................................5
2 Theoretical background of the resistance spot welding process ........................8
2.1 History of resistance spot welding ..............................................................8
2.2 Physical phenomena in resistance spot welding .........................................8
2.3 Resistance spot welding parameters ........................................................12
2.4 Resistance spot welding equipment .........................................................17
3 Physically based process planning of resistance spot welding .........................20
3.1 Evaluation criteria of resistance spot welding ..........................................20
3.2 Physical testing procedure ........................................................................22
4 Model based process planning of resistance spot welding...............................26
4.1 Regression models of RSW ........................................................................26
4.2 Finite element modeling for RSW simulations ..........................................27
5 Results and discussions .....................................................................................33
5.1 Variations of resistance spot welding .......................................................33
5.2 Regression models of resistance spot welding .........................................35
5.3 Material models for simulations of resistance spot welding ....................40
5.4 Physical verification of resistance spot welding FE simulations ...............47
6 Conclusions and future research .......................................................................50
6.1 Conclusions................................................................................................50
6.2 Future work ...............................................................................................50
7 References .........................................................................................................52
iii
Nomenclature
B Flow stress parameter [-]
H Enthalpy [J]
I Current [A]
Q Heat [W]
q Pressure [Pa]
t Time [s]
T Temperature [K]
u Displacement [m]
V Voltage [V]
z z-direction [-]
ε Strain [-]
ρ Density [kg·m−3]
iv
κ Electrical conductivity [S m-1]
σ Stress [Pa]
v
vi
1 Introduction
Industries and consumers in general are subject to pressure to reduce greenhouse
gas emissions, caused by their manufacturing and the use of their products, to
reduce the effects of global warming [1] The automotive industry – one of the
dominant sources of greenhouse emissions [2] - has answered to this pressure
mainly by reducing fuel consumption by producing more fuel efficient engines,
reducing the weight of products [3] and introducing alternative fuel engines [4].
The reduction in weight of vehicles is made possible by innovative designs and use
of materials of high strength to weight ratios, such as high-strength steels,
aluminum or composite materials [5]. These solutions are associated with
challenges in manufacturing due to the novelty and complexity of the components
and materials.
1
Materials
Equipment Weld
Process
parameters results
Application
1.1 Background
In the transportation industry, and the automotive industry in particular, resistance
spot welding (RSW) has for decades been the main joining technology. The
method’s low cost, high reliability, high time efficiency, high accessibility and high
ability for robot automatization, compared to other joining methods, make it ideal
for automotive production. A typical modern automobile contains around 3000 –
5000 resistance spot welds, while other joining methods, such as laser welding, arc
welding or mechanical fastenings is used in a much more limited scale. A typical
resistance spot weld is shown in Figure 2. [6]
The large scale of RSW in manufacturing makes the method’s results highly
important for product properties, including critical properties such as vehicle
safety, crashworthiness and fuel efficiency [6]. In order to ensure robustness in
RSW applications, variations of process results should be minimized. In order to be
minimized, the process variations must be correctly understood and predicted. An
important step for understanding and predicting process variations is through
process planning.
2
Figure 2 Section cut of resistance spot weld between two sheets.
To measure weld quality and process robustness of RSW, the main measurements
are weld nugget size (the lateral size of the solidified joint zone) and the occurrence
of weld expulsion. Expulsion is the phenomenon which occurs when molten metal
is ejected from the weld. While, the two measurements do not fully measure all
aspects of weld quality, they indicate several important factors and are relatively
easy to attain.
In general, process planning may include a broad range of analysis and measures.
Regarding RSW, however, process planning is focused on process parameters. By
understanding the effect of process parameters and their optimization, the RSW
process’ result properties and robustness can be improved.
The lack of accurate large scale non-destructive test (NDT) methods for RSW
increases the importance of process planning for RSW applications [7]. Therefore,
the predictive capability of process planning tools must be sufficient in order to
accurately model production scale RSW processes.
3
advanced FE simulations [9] have been used to model welding processes
accurately. Also, specifically regarding RSW, numerical methods have been shown
in literature but a large-scale use of simulations for process planning is yet to be
presented. A more detailed review of the development of numerical models is
presented in this thesis.
In this thesis, virtual process planning of RSW is treated through several methods.
Studies showing the variations of RSW results as well as the capability of numerical
methods for large scale implementation are presented. Such variations have been
previously reported [10], [11] but have not been quantified. The variations are
studied through repeated welding of apparently identical welds and destructively
tested and evaluated. The numerical methods are studied through development of
regression models and FE simulations. The results of the numerical models are
compared to physical tests, which are destructively tested and evaluated. A more
detailed introduction to the appended papers is made in chapter 1.5.
The first question regards variations of RSW properties and is stated as follows.
4
II: How accurately can numerical models predict resistance spot weld
sizes?
As an initial study of process planning, the variations of RSW results were analyzed.
The first paper includes experimental RSW results from a laboratory environment,
where apparently identical welding was performed repeatedly. A large amount of
weld size data was recorded and the variations in weld sizes were measured and
analyzed. The experiments were done to quantify the variations of RSW results in
apparently identical welding conditions, due to natural variations.
The effects of electrode tip dressing, welding current, electrode force and contact
resistance on weld nugget size and expulsion occurrence, respectively, were
analyzed.
5
Furthermore, statistical modeling to describe the result variations was performed.
Such a statistical model, in form of a distribution function, can further provide
understanding of the variations of the process.
The paper provides analyses of causes of variations and a statistical model for
describing variations and is intended to answer research question I.
The second paper is a continuation of the first paper. Here, RSW results from
laboratory as well as production environments were included.
Further measurement and analysis of weld results were presented. The paper
aimed to show the possible differences in results between the controlled, ideal
conditions of the laboratory and the production environment where the RSW
process is affected by several other aspects of manufacturing variations and more
complex geometries. The paper gives a more detailed answer to research question
I.
The third paper used experimental welding results from a broad range of material
configurations as base for regression models used for RSW process planning.
Several regression models were generated and evaluated through systematic
statistical methods, as an answer to research question II. The regression models
were evaluated in terms of requirements on input data and ability for industrial
implementation as well as prediction capability.
The models show the effect of process parameters and material configuration on
RSW results. The models can thus be used for parameter optimization as well as
process robustness studies.
6
1.5.4 PAPER D: Verification of the Capability of Resistance Spot
Welding Simulation
In Proceedings of the Sheet Metal Welding Conference XV, 2012.
The fifth paper is a continuation of the fourth. The material range for verification
has been expanded to include six materials of varying strength and material models
have been generated for all materials. Furthermore, a more detailed analysis of
expulsion has been performed in the virtual process planning environment.
The paper aims to further show the capabilities of simulations of RSW and works as
a more general answer to research question II.
7
2 Theoretical background of the
resistance spot welding process
In order to form a framework of the thesis, the RSW process is presented and
described in the following chapter.
The principle behind the joining method is to pass a current through two or more
sheets as they are clamped together by an electrode pair. Furthermore, the history,
advantages, disadvantages and common applications of spot welding are described
in the following chapters.
(1)
The heat generation rises the temperature of the materials above the melting point
mainly due to surface resistances between the sheets. After the metal sheets have
molten the current is stopped and no further heat is generated. The temperature
8
will decrease again and the joined sheets form a solid nugget, joining the sheets.
The outline of the spot welding process is illustrated in Figure 3.
The heat generated due to the electrical current is defined through Joule's law,
equation 1. The resistances, and thus heat generation, of the circuit are located at
four major sources, see Figure 4: Typical values of resistances of materials used in
RSW can be seen in Figure 5.
For stack-ups with more than two sheets, bulk resistances and contact resistances
are added accordingly.
9
Figure 4 Cross-section of spot weld process and electrical resistances (1-7)
10
Since the resistances are connected in a series electrical circuit they will all retard
current flow. However, the resistance is a highly temperature dependent variable
and will change significantly during the weld process. It can be noted that the
resistivity of both aluminum and steel is larger than that of copper of the
electrodes. Thus, the heat generated in the work piece is greater than that in the
electrodes, which enables welding. This also explains the difficulties to weld
aluminum sheets compared to steel, due to the smaller heat generation. However,
the contact resistances between the materials are significantly larger and thus,
determine the heat generation more, than the bulk resistances of the materials. [6]
To assure electrode contact with the work piece, the electrodes are clamped to the
sheets with a mechanical force. The electrodes are most commonly made of a
copper alloy. When the electrodes and work piece are in full contact, the
electrodes are charged and high amperage current is flowing through the circuit.
After the current flow stops the electrodes are squeezing the sheets during a
further period, called cool time, to improve controlled nugget formation. The
outline of the spot welding process is illustrated in Figure 3. The influence and
effect of the input parameters are described in chapter 2.3
For non-constant current or resistance, which is often the case in RSW, the
expression in Equation 1 must be integrated to find the correct heat energy. The
formula gives a basic understanding of the circuit but is not detailed enough for full
understanding. Firstly, in resistance spot welding the local heat generation at each
resistance is of more interest than the total heat energy of Joule's law. Secondly,
the heating rate is a more indicative parameter of weld development than total
heat energy since it has a significant effect on the microstructures of the weld.
All metals involved in the resistance spot welding process increases in electrical
resistance with temperature. Steel is very sensitive to temperature changes, the
difference between steel and copper explain the excessive heat in the steel sheets
compared to the heat in the electrodes. In addition, the electrodes are water-
cooled which decreases their temperature increase, and resistance, further.
11
Another electrical phenomenon that must be taken into account is shunting
currents. Such currents occur when a weld is made in close relation to previously
made welds. The adjacent welds create a current path which distorts the intended
welding by lowering the effective welding current or current density.
12
current-force figure as shown below. In the following chapter each of the machine
parameters and their impact on the spot weld is discussed.
As with weld current level a too short weld time will not generate sufficient heat to
form a weld a too short weld time may lead to over-heating or expulsion.
13
The surface deformations are especially important in areas where visual quality is
of high importance. Examples of such areas may be outer sides of car bodies.
On the other hand, a too low force will increase the risk of geometrical instability of
the welding process and excessive heat generation. In other words, the risk for
expulsion is increased with a lower force.
The tip curvature of the electrode is a measure against the degradation of the
electrode tip. As the electrode degrades the initial curvature will deform into a flat
surface. An initially flat surface would cause a concave tip after continuous welding.
The most commonly used electrode geometries are described in standard ISO-
5821:2009 [15].
14
Figure 7 Electrode geometry types
15
degradation mechanisms. The uneven tip will worsen weld conditions and increase
risk of expulsion at the sheet-electrode interface. A heavily degraded electrode tip
is shown in Figure 8.
Pre-pulsing is technique were a short duration, high current is passed through the
materials before the actual welding current. Such a pre-pulse enables galvanizing
16
layers to be removed to enable better welding results during the actual weld time
by improving sheet-electrode contact. It is also reported [18] to enable successful
welding of more challenging sheet combinations, such as uneven thicknesses of
multi-sheet stack-ups.
Another design aspect is the kinetics of the electrodes. Two main principles are
used: pneumatic and electric guns. Electric guns are generally more powerful, more
exact and more efficient but may be of a higher cost compared to pneumatically
moving electrodes.
Also, the mechanics of the weld gun arms can differ. Two common alternatives are
dominating, linearly moving electrodes, see Figure 9, and so called X-guns where
the two arms are rotating along a common axis. An X-gun transfers the momentum
at the axis to an electrode force at the electrode tips. Linear guns need less space
and provide a pure axial force to the metal sheets but may also be more expensive
compared to X-guns.
17
Figure 9 RSW linear gun attached to a robot.
The reasons behind the development towards MFDC power sources are both
increasing weld size and energy efficiency. Two identical set-ups with equal root-
mean-square currents using AC and DC produce larger welds in DC mode, especially
at lower currents. It is suggested that DC power sources create a larger contact
resistance and thus more heat. Another reason is the electrical inductance caused
18
by the alternating current. In a case study comparing AC and MFDC power sources,
it is suggested that the energy efficiency of MFDC is approximately 42% higher than
AC power sources. In other words, 42% less energy is lost in the equipment. [19]
19
3 Physically based process planning
of resistance spot welding
Process planning of resistance spot welding focuses on assuring of weld properties
of process parameters before industrial production is started. In order to assure
quality of established production and to examine new solutions, completed
resistance spot welds are regularly tested and evaluated. Due to the uncertainties
and largely automated and unmonitored production of spot welds, there is a great
need for testing and verification of weld quality in the industry prior to large scale
manufacturing.
Traditionally process planning has been performed through physical testing of weld
properties. Physical testing of resistance spot weld tests can be divided into two
categories: experimental tests under controlled environments and tests from
industrial production. The experimental tests are often performed to examine a
material configuration’s weldability whereas production tests are made to assure
weld quality of a configuration already in production.
In the following chapter, only weld geometries have been treated, i.e. tests of
strength capacities or hardness tests have not been directly treated.
20
Figure 10 A typical cross-section of a spot weld and evaluation parameters. [6]
Of the parameters listed above, nugget size is the most indicative of weld strength
and thus most frequently measured [6]. However, it is not sufficient to describe
weld quality by itself and for a complete understanding of the properties of a weld
several parameters needs to be measured. For example, penetration depth
indicates the volume of material which melted during the weld process. Insufficient
penetration indicates too low heat input while excessive penetration results in
softening of the material and large indentation. Another, indication of weld
penetration is the magnitude of the process margin. While, a larger penetration
does not automatically result in higher strength, it shows that certain variations of
weld output may be allowed with maintained results.
21
Figure 11 Weld size requirements from different organizations [6]
A similar comparison for penetration depth shows that requirement varies from
less than 25% penetration of sheet thickness up to a minimum of 80% for certain
military applications of aluminum welding.
22
Labs may conduct both experimental and production tests. Both are visually
examined through metallographic testing to gain weld outcome. Hardness and
mechanical strength tests can also be done to gather full understanding.
Production tests are often, as mentioned earlier, samples from production and the
welding process is not performed and controlled in a lab environment.
The experimental tests, on the other hand, are carried out on small coupons of
standardized geometry. For each material configuration the weld current is step-
wise increased for each coupon with two welds on each coupon. The double welds
are performed in order to account for interfering shunt currents - the second weld
is examined.
23
Figure 13 Test coupon after coach peel test
The weld cross-section is then optically examined by microscopy and the relevant
parameters are obtained and recorded by computer.
The main disadvantages and uncertainties involved in the described testing method
are those involved with in using a 2-d test surface to describe the properties of a 3-
d weld volume. A center cross-section may not show all properties of the entire
weld, such as porosity, voids or other imperfections.
24
Figure 14 A typical 1-d weld lobe showing the accepted weld current range.
A 1-d weld lobe is a concise illustration of the weldability of a material
configuration, making it ideal for many industry purposes. 2-d weld lobes in which
both current and weld time or electrode force have been varied can also be used.
25
4 Model based process planning of
resistance spot welding
In order to replace physical based process planning of RSW, several models can be
used. In the present thesis regression models and finite element models have been
studied. The two modeling approaches are treated in the present chapter.
(2)
Regression models of the RSW process have to some degree been subject to
research [21], [22], [23]. However, these models only treat only one material each
and are thus not ready for a larger scale industrial implementation.
The regression model is generated through a least-square fitting, by the known set
of input data and results, (X, Y) as shown in equation (3) [24].
(3)
In in equation (3), the matrix X includes the known factors and the matrix Y
includes the known results.
In RSW process planning, the most important measurement is weld nugget size and
its sensitivity to process parameters. Thus, the regression model should describe
the effects of changes in process parameters to weld nugget size. The benefits of a
regression model is the mathematically simplicity of the numerical model, while
having high predictive capability if generated correctly.
The factors known to affect the weld nugget size are many and are often
dependent of each other. The factors included in the regression model should
satisfy several qualities; they should be physically and statistically significant while
being easily accessible for process planning.
The effect of different factors on the weld size and thus their relevance in the
regression model can be evaluated by ANOVA (ANalysis Of VAriance) [24]. ANOVA
26
is a set of statistical tests used for evaluation of data. For the present work, two
ANOVA tests have been used; the F-test and the t-test. The F-test is designed to
evaluate the significance of all factors of the model in one test. The t-test is
designed to evaluate the individual factors of a test. In order to gain a statistically
significant both tests should give successful test results.
The first research work in RSW modeling [25] focused on one-dimensional models
of the temperature distribution of the RSW process. Later the first axi-symmetrical
model was developed [26]. Later axi-symmetrical models also measured the
contact diameter between the sheets [27]. Research continued with a one-
dimensional model which took the temperature dependence of resistance into
account [28]. Other axi-symmetrical models took the contact pressure dependence
of contact resistance [29] or the steel hardness dependence of contact resistance
[30] into account. A one-dimensional model was developed which measured weld
size through temperature dependent contact resistance and material properties
[31]. These earlier models were based on the Finite Difference Method, which
limited their accuracy.
Development continued with models based on the Finite Element Method with
constant material parameters and a fully elastic mechanical model [32] and an
elasto-plastic model [33], [34], [35]. Later, coupling of the thermal, electrical and
mechanical fields was introduced in simulations [36], [37]. Also, research was
published which focused on 3d-models of RSW [38]. Models were developed with
more detailed analysis of contact resistance conditions [39], [40] [41]. Work was
focused on modeling of steel sheets, but research treating aluminum sheets was
also introduced [42]. Due to the large and rapid temperature variations during the
RSW process, metallurgical modeling was later introduced [43],
27
RSW simulations. In the present thesis, Sysweld and Sorpas have been studied and
used and the modeling is based on these software.
28
4.2.4 Modeling and boundary conditions
In the simulations, a number of principal constitutive relations control the outcome
of the simulations. These consecutive relations are often coupled but may still be
defined individually. In the below section, the constitutive relations which control
the simulations are treated. They have been grouped into thermal, electrical and
mechanical modeling principles. In addition, each governing equation is controlled
by a set of boundary conditions, presented in chapter 4.2.8.
(4)
(5)
In tandem with the principle of conservation of current flux, see equation (6), the
above equations form the base of the electrical model.
(6)
(7)
The first term includes the enthalpy and the density of the system. The second
term includes the conductive heat and the third term is the Joule heating term. The
fourth term includes the internal heat generation.
29
are not included in the model due to their small effect. Mathematically expressed,
the term Q is neglible.
Another aspect worth noting is the distinction between enthalpy and specific heat -
both properties describe the thermal state of the body. However, to fully take
phase transformation effects into account accurately, the enthalpy model is more
convenient.
(8)
(9)
(10)
The matrices [D] and [C] describe the elastic and thermoelasto-plastic states,
respectively.
30
In parallel with the governing equations, the model is based on a set of boundary
conditions for the thermal, electrical and mechanical models. Each is treated
separately below.
Water channel
Symmetry line
Metal-air
sheet
interface
31
Heat loss at electrode-water interface.
(16)
n is the normal direction of the boundary. The function K(T) is a combined function
of the convection heat transfer coefficient and the radiation heat transfer
coefficient.
T is the temperature of the surrounding medium, i.e. the air temperature. In the
Sysweld model, K(T) includes both convective and radiation heat transfer closer to
the electrodes but only convective heat transfer further away from the electrodes.
At symmetry lines.
(18)
32
5 Results and discussions
To answer the research questions in chapter 1.3 a number of studies have been
performed. The results from these studies and answers to the research questions
are summarized in the following chapter. A more detailed description of the studies
is included in the appended papers.
An analysis was performed to quantify and analyze the variations of weld quality in
RSW applications. As a collection of results for analysis, controlled laboratory
welding and production welding were carried out. Several apparently identical
welds were used for analysis.
The laboratory trials were welded with two coupons of rephosphorized steel, a
common material in automotive production. For the production welds, three joints,
which all included two sheets of rephosphorized steel sheets, were selected.
All welds were destructively tested using coach peel tests and weld nugget sizes
were measured and recorded. The results of typical standard deviations of the
series from laboratory and production environments are shown in Figure 17.
33
1.00
Standard deviation [mm]
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
Laboratory conditions Production conditions
34
Figure 18 Distributions of typical RSW results.
The results of the present studies can be used to draw a number of conclusions
about variations of weld nugget. Typical standard deviations of resistance spot
weld nugget sizes in laboratory environments are approximately 0.30 mm in
magnitude. Typical standard deviations of resistance spot weld nugget sizes in
production are approximately 0.90 mm in magnitude. Moreover, results show that
both fitted normal and Weibull distributions show promising results in describing
weld size distributions.
35
been generated. The input for the models is results from physical testing, which has
been carried out for an extensive range of materials and equipment in modern
RSW applications. In total, over 200 spot weld results have been used for analysis.
The models predict weld nugget diameter, which, as described in chapter 3.1 is the
most common measure of weld quality. In the analysis weld parameters such as
weld current and electrode force, equipment parameters such as electrode tip area
and material parameters such as material thickness. The distributions of weld
parameters are shown in Figure 19. Both linear and quadratic terms have been
used for the analysis.
36
Figure 19 Distributions of input parameters for regression models.
Three different regression models were generated using iterative step-wise
regression methods. The methodology of the iterative step-wise regression is
shown in Figure 20 and is based on Analysis of Variance (ANOVA) and adjusted R2-
value [24]. The first model included pre-selected factors which were forced to be
37
included in the model. The second and third model included linear and quadratic
terms, respectively, without any pre-selected factors.
All models gave successful
results in terms of ANOVA
tests, which indicate the
models are statistically
significant. However, the
models gave somewhat
different results in terms of
factors included in the
model, despite the
successful ANOVA tests.
The various selected factors
of the models indicate the
complex behaviour of weld
growth during RSW. It also
indicates the dependency of
many physical phenomena
in weld nugget growth.
The forced, or pre-selected
model, resulted in the
highest R2-value, thus
indicating the highest
predictive capability.
Secondly, the linear model
showed the lowest
predictive capability
indicating that linear terms
do not capture the weld
results to as high degree.
The quadratic model
showed intermediate
predictive capability.
Figure 20 Step-wise regression methodology However, the quadratic
model has the lowest
number of factors compared to the other models, which may be advantageous in
engineering use. The standard deviation of the error, which also is a measure of the
predictive capability is similarly ordered; the forced model has the lowest standard
deviation (0.5 mm) followed by the quadratic (0.6 mm) and the linear model (0.7
mm). The regression model results compared to the physical results are shown in
Figure 21, Figure 22 and Figure 23, respectively.
38
Figure 21 Regression accuracy of forced model.
39
Figure 23 Regression accuracy of quadratic model.
In the following part, the material data for the selected materials are presented.
The method of gathering the data and data evaluation is also presented through
standard deviations compared to experimental data where relevant. In some cases
generalization have been made, which are also presented.
40
high strength steels, such as boron steel or martensitic steels, to lower strength
steels, such as Dogal 260, which is a rephosphorized steel.
From the results from IDS linear or second order regression models were built for
general expressions of thermal conductivity for all materials and temperatures. A
common model for base material and austenite was used.
41
Figure 24 Thermal conductivity model.
For the simulations, linear interpolation was used between the data points
obtained from ThermoCalc. A common model for base material and austenite was
used.
42
Figure 25 Specific heat model.
Since the deviations between materials were small a general model was used for
simulation for each microstructure. Also, the thermal expansion coefficient is
considered to be independent of temperature. The coefficient of thermal
expansion used in simulation was 1.38e-5 for the base material and 1.96e-5 for
austenite.
In order to obtain the material data, several specimens of each material were
tensile tested at different temperatures and micro-structures. Tensile testing was
performed to gather the engineering stress and strain. The raw data is test
43
specimen specific and not material general and must be converted into true stress
and strain for model generation.
For the simulations, a common Young’s modulus model has been used for all
materials and all micro-structures. A linear approximation, as seen in Figure 26 has
been used. The linear approximation is valid up to a threshold value, which is 1% of
the value at room temperature. At temperatures above the threshold value, the
Young’s modulus has been modeled as constant. The standard deviation between
the model and the experimental results was 26.7 GPa.
44
5.3.3.3 Yield stress
The yield stress determines the transition between elastic and plastic behavior in
the mechanical model. The processing of tensile test data to obtain the yield stress
used the Young’s modulus approximation, described in Chapter 5.3.3.1. The
Young’s modulus will create a linear stress-strain curve, which in the elastic region
is near identical to the test stress-strain curve. However, as the plastic
deformations take place, the perfect linear curve and the test curve will start to
deviate. A conventional method of defining a yield stress is by locating a specific
divergence between the two curves, i.e. a specific plastic strain magnitude. In the
present work, the yield stress is defined as the stress where the plastic strain is 0.2
%.
For yield stress, several approaches were evaluated to generate a general model
for simulations; a linear interpolation model, a linear regression model similar to
the Young’s modulus model and Johnson-Cook models [49]. It was concluded that
the linear regression model was the most accurate in modeling yield stress. The
yield stress models are shown in Figure 27. The standard deviation between the
model and the experimental results was between 60.8 MPa for the DP600 material
and 137.0 MPa for the martensitic steel.
45
5.3.3.4 Flow stress
The flow stress describes the plastic behavior, i.e. the deformations above the yield
stress limit. The true test stress-strain curves from the tensile tests have been used
to establish a general material model for the plastic behavior for simulation. To
establish the general model for all temperatures both Johnson-Cook models and a
parameter optimization were evaluated.
The parameter optimized stress-strain relation showed very high correlation to the
test data while the Johnson-Cook models were found to be insufficient. The
optimization method used MATLAB Optimization Toolbox [50] to fit the parameters
B and n in Equation (23) to the tensile test results..
(23)
After the optimization, the standard deviation between the model and the
measured stress-strain curve was between 0.66 MPa and 79.15 MPa. The flow
stress curves are seen in Figure 28. Additionally, no deformation hardening took
place at temperatures above the melting temperature
46
5.3.4 Other material data
In resistance heating, the electrical properties of the sheet materials are of high
importance. For the electrical model, the electrical conductance is used as input
data. The electrical conductance will control the Joule heating of sheets in the
simulation. The electrical conductivity is controlled by a master curve, which is
common for all materials and phases [45].
For the contact conditions, a element layer of specific thermal and electric contact
conductances was included in the model. The conductances used a general model
for all materials and were dependent of temperature. [45]
47
The physical experiments generated weld nugget sizes for each current and
material and the occurrence of expulsion. The welding was performed on smaller
sheet coupons and two welds were welded on each coupon stack-up. The two
welds would analyze the effect of shunting currents to the previous welds.
The FE simulations calculate the weld nugget size as the size of the molten zone at
the interface at the end of the weld time. Also, the simulations gave an expulsion
limit based on the weld zones contact with the air gap. The simulation does not
take shunting into regard
A comparison between weld sizes from simulation and physical welding can be
seen in Figure 30. The standard deviation of the residual between the weld size
diameter of the FE model and the experiments is 0.68 for all welds. The first weld is
more accurately modeled with a standard deviation of 0.54 mm compared to 0.80
for the second weld. As seen in chapter 5.1 standard deviation in experimental
welding is approximately 0.3 mm.
The comparison shows that the standard deviation of the residual of the expulsion
limit is 1.27 kA. The values for the first and second weld location are 1.15 kA and
1.27 kA, respectively. I.e., the expulsion limit better capture the expulsion of the
first weld location.
Expulsion limits has variations in physical testing just like weld sizes. The standard
deviation of the expulsion limit in physical testing can be measured by the discrete
probability distribution. The standard deviation of expulsion limit is 0.21 kA
considering both weld locations and all materials.
The present results show the capability of RSW FE simulations. Compared to many
other comparable studies, the present work has an extensive scope of physical
experimental results. The extent of the comparison shows the benefit of FE
simulations for industrial implementations.
48
Figure 30 Results from simulations and physical testing of weld nugget size for all
materials of welds from the first weld location (blue) and second weld location
(green).
49
6 Conclusions and future research
The following chapter shows the main conclusions from the present thesis and
presents directions for future research work.
6.1 Conclusions
The present thesis focuses on process planning of resistance spot welding. A
number of studies have been carried out to answer research questions regarding
variations of spot weld sizes and predictive capability of numerical methods of the
RSW process. A number of conclusions can be drawn from the studies.
Firstly, variations of RSW weld sizes were statistically analyzed. The studies showed
that under apparently identical conditions, weld sizes vary by a significant
magnitude. The magnitude of variations was also seen to be different between
controlled laboratory conditions and production conditions. The standard deviation
of laboratory welds and production welds were approximately 0.3 mm and 0.9 mm,
respectively.
Also, the studies of weld size variations showed the distribution of weld sizes. It
was seen that both in laboratory and production conditions, both normal and
Weibull distributions could describe the distribution of weld sizes.
Further studies analyzed numerical methods of the RSW process by both regression
models and FE simulations. The numerical methods were evaluated by comparison
to physical welds. Both models showed promising results by physical verification.
The standard deviations of the regression models were in the order of 0.5 mm to
0.7 mm and for the regression models and in the order of 0.7 mm for the
simulations
For the FE simulations, material models for several materials were generated. The
models included thermal, mechanical and electrical data for elevated
temperatures. The models were generated by means of material simulation,
material testing and literature studies.
The first part of the thesis included analysis of variations of RSW weld sizes. A key
conclusion form the results was the difference in magnitude of variations between
50
laboratory and production conditions. While, variation in spot weld location,
through shunting currents, were seen to be a source for the difference, other
sources are possible. To further minimize variations, it is of interest to establish and
quantify other disturbance sources, such as geometrical variations and variations in
weld equipment condition
Regarding the numerical models, both regression and FE models, the results
showed their predictive capability. While, the verification was performed on 2-
sheet joints of similar materials, current industrial application often include joints
of more than two sheets and often of dissimilar materials and thicknesses. In order
to fully evaluate the capability of the methods, it is interest to verify with a larger
scope of different joints.
51
7 References
[1] S. Oberthür and H. Ott, The Kyoto protocol, Berlin [u.a.]: Springer, 1999.
[4] A. Topolansky, "Alternative fuels: Challenges and opportunities for the global
automotive industry," The Columbia Journal of World Business, vol. 28, no. 4,
pp. 38-47, 1993.
52
[11] M. Galler and W. Ernst, "The influence of electrode wear on current density
IIW SC 27 09," IIW, 2009.
[12] P. Houldcroft, Feng, Fim and Fweldi, "Welding process developments and
future trends," Materials & Design, vol. 7, no. 4, pp. 162-169, 1986.
[14] G. Solas, J. Sharp and H. Goldsmid, Eds., Thermoelectrics. Basic Principles and
New Materials Developments, Springer Heidelberg / New York, 2001.
[16] ISO 5182:2008 Resistance welding -- Materials for electrodes and ancillary
equipment.
[18] I. Khan, M. Kuntz, L. Zhou, Y. Chan and N. K. Scotchmer, "Monitoring the Effect
of RSW Pulsing on AHSS using FEA (SORPAS) Software," vol. 2139, pp. 37-44,
2007.
[20] C. Ma, D. Chen, S. Bhole, G. Boudreau, A. Lee and E. Biro, "Microstructure and
fracture characteristics," Materials Science and Engineering A, vol. 485, p. 334–
346, 2008.
[22] Y. Luo, J. Liu, H. Xu, C. Xiong and L. Liu, "Regression modeling and process
analysis of resistance spot welding on galvanized steel sheet," Materials
53
\& Design, vol. 30, no. 7, pp. 2547-2555, 2009.
[23] M. Zhou, H. Zhang and S. Hu, "Relationships between Quality and Attributes of
Spot Welds," Welding Journal, vol. April, pp. 72s - 77s, 2003.
[24] H. Sahai and M. Ageel, The Analysis of Variance: Fixed, Random, and Mixed
Models, Birkhäuser, 2000.
[27] Y. T. and T. Okudo, "A Study of Spot Welding of Heavy Gauge Mild Steel,"
Weldíng in the World, vol. 9, pp. 234-255, 1971.
[30] H. Cho and Y. Cho, "A study of the thermal behaviour in resistance spot welds,"
Welding Journal, vol. 68, no. 6, pp. 236s-244s, 1989.
[32] H. Nied, "The Finite Element Modelling of the Resistance Spot Welding
Process," Welding Journal, vol. 63, pp. 123s-132s, 1984.
[34] P. S. Wei and C. Y. Ho, "Axisymmetric Nugget Growth During Resistance Spot
Welding," Journal of Heat Transfer, vol. 112, no. 2, pp. 309-316, 1990.
54
[35] P. S. Wei and F. B. Yeh, "Factors Affecting Nugget Growth With Mushy-Zone
Phase Change During Resistance Spot Welding," Journal of Heat Transfer, vol.
113, no. 3, pp. 643-649, 1991.
[37] O. P. Gupta and A. De, "An Improved Numerical Modeling for Resistance Spot
Welding Process and Its Experimental Verification," J. Manuf. Sci. Eng., vol.
120, pp. 246-252, 1998.
[41] L. Xu and J. Khan, "Nugget growth model for aluminum alloys during resistance
spot welding," Welding Journal, vol. 78, p. 367s–372s, 1999.
55
[46] W. Zhang, "Design and implementation of software for resistance welding
process simulations," Journal of Material and Manufacture, vol. 112, p. 556–
564, 2003.
[48] U. Dilthey and P. Ohse, "The front line in modeling of resistance welding,"
Journal of Japan Welding Society, vol. 76, p. 14–18, 2007.
[49] G. R. Johnson and W. H. Cook, "A constitutive model and data for metals
subjected to large strains, high strain rates and high temperatures," 1983.
56